Increasing the accuracy of brain functional maps through large deformation diffeomorphic metric mapping

The accuracy of the brain normalization method directly impacts the preciseness of statistical analysis of functional magnetic resonance imaging (fMRI) data. Furthermore, the study of the medial temporal lobe and cortical layer structures requires an accurate co-registration method due to large inter-subject variability. In this thesis, we first introduce a fully automated fMRI post-processing pipeline aimed to reduce the registration error during group studies and we will demonstrate its superiority over two widely used registration methods by conducting a comprehensive bleeding study using a synthesized fMRI data-set as well as surface-to-surface distance quantifications over both cortical and sub-cortical regions. Finally, we apply our processing pipeline to a functional MRI data-set of a schizophrenia study and show how accurate registration of hippocampus and inferior frontal gyrus structures can increase the accuracy of functional maps over these regions when performing group analysis.